Indirect measurement of railway pantograph contact force using model-based and data-driven approaches in laboratory tests

The contact force between the pantograph and the overhead contact line is a key indicator of railway current collection performance. Conventional measurement techniques rely on instrumented pantographs with structural modifications, increasing system complexity and limiting practical implementation. This work proposes and experimentally validates a low-intrusive methodology to estimate the pantograph contact force using acceleration, displacement, and pressure measurements while avoiding force measurements. The estimation problem is separated into two frequency ranges. The very low-frequency component is reconstructed from the pressure signal of the pantograph uplift mechanism, while the remaining frequency content is obtained by solving an inverse dynamic problem based on displacement and acceleration measurements. Four estimation models are investigated: two model-based approaches (Kalman Filter and Moving Horizon Estimation) and two data-driven models (feedforward Neural Network and Long Short-Term Memory network). Model parameters and hyperparameters are optimised to ensure a fair comparison and fully exploit the models’ potential. The methodology is experimentally validated on a laboratory test bench with a real pantograph. Results show reconstruction of the contact force within the 0–20 Hz range defined by EN 50317.